Learning from Memory: Non-Parametric Memory Augmented Self-Supervised Learning of Visual Features

Thalles Silva, Helio Pedrini, Adı́n Ramı́rez Rivera
Proceedings of the 41st International Conference on Machine Learning, PMLR 235:45451-45467, 2024.

Abstract

This paper introduces a novel approach to improving the training stability of self-supervised learning (SSL) methods by leveraging a non-parametric memory of seen concepts. The proposed method involves augmenting a neural network with a memory component to stochastically compare current image views with previously encountered concepts. Additionally, we introduce stochastic memory blocks to regularize training and enforce consistency between image views. We extensively benchmark our method on many vision tasks, such as linear probing, transfer learning, few-shot classification, and image retrieval on many datasets. The experimental results consolidate the effectiveness of the proposed approach in achieving stable SSL training without additional regularizers while learning highly transferable representations and requiring less computing time and resources.

Cite this Paper

BibTeX
@InProceedings{pmlr-v235-silva24c, title = {Learning from Memory: Non-Parametric Memory Augmented Self-Supervised Learning of Visual Features}, author = {Silva, Thalles and Pedrini, Helio and Ram\'{\i}rez Rivera, Ad\'{\i}n}, booktitle = {Proceedings of the 41st International Conference on Machine Learning}, pages = {45451--45467}, year = {2024}, editor = {Salakhutdinov, Ruslan and Kolter, Zico and Heller, Katherine and Weller, Adrian and Oliver, Nuria and Scarlett, Jonathan and Berkenkamp, Felix}, volume = {235}, series = {Proceedings of Machine Learning Research}, month = {21--27 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v235/main/assets/silva24c/silva24c.pdf}, url = {https://proceedings.mlr.press/v235/silva24c.html}, abstract = {This paper introduces a novel approach to improving the training stability of self-supervised learning (SSL) methods by leveraging a non-parametric memory of seen concepts. The proposed method involves augmenting a neural network with a memory component to stochastically compare current image views with previously encountered concepts. Additionally, we introduce stochastic memory blocks to regularize training and enforce consistency between image views. We extensively benchmark our method on many vision tasks, such as linear probing, transfer learning, few-shot classification, and image retrieval on many datasets. The experimental results consolidate the effectiveness of the proposed approach in achieving stable SSL training without additional regularizers while learning highly transferable representations and requiring less computing time and resources.} }
Endnote
%0 Conference Paper %T Learning from Memory: Non-Parametric Memory Augmented Self-Supervised Learning of Visual Features %A Thalles Silva %A Helio Pedrini %A Adı́n Ramı́rez Rivera %B Proceedings of the 41st International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2024 %E Ruslan Salakhutdinov %E Zico Kolter %E Katherine Heller %E Adrian Weller %E Nuria Oliver %E Jonathan Scarlett %E Felix Berkenkamp %F pmlr-v235-silva24c %I PMLR %P 45451--45467 %U https://proceedings.mlr.press/v235/silva24c.html %V 235 %X This paper introduces a novel approach to improving the training stability of self-supervised learning (SSL) methods by leveraging a non-parametric memory of seen concepts. The proposed method involves augmenting a neural network with a memory component to stochastically compare current image views with previously encountered concepts. Additionally, we introduce stochastic memory blocks to regularize training and enforce consistency between image views. We extensively benchmark our method on many vision tasks, such as linear probing, transfer learning, few-shot classification, and image retrieval on many datasets. The experimental results consolidate the effectiveness of the proposed approach in achieving stable SSL training without additional regularizers while learning highly transferable representations and requiring less computing time and resources.
APA
Silva, T., Pedrini, H. & Ramı́rez Rivera, A.. (2024). Learning from Memory: Non-Parametric Memory Augmented Self-Supervised Learning of Visual Features. Proceedings of the 41st International Conference on Machine Learning, in Proceedings of Machine Learning Research 235:45451-45467 Available from https://proceedings.mlr.press/v235/silva24c.html.

Related Material